OpenFace: A general-purpose face recognition library with mobile applications

By Brandon Amos, Bartosz Ludwiczuk, Mahadev Satyanarayanan. 2016

1,082 citations recorded by the source corpus.

Subfields: Computer Vision and Pattern Recognition

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Computer Vision and Pattern Recognition

Citations in the corpus

FairFace: Face Attribute Dataset for Balanced Race, Gender, and Age (2019, background): “raphic survey performed in marketing or social science research, aimed at understanding human social behaviors and their relations to demographic backgrounds of individuals. Using off-the-shelf tools [2,4] and commercial services, social scientists, who traditionally didn’t use images, begun to use ”
MoCoGAN: Decomposing Motion and Content for Video Generation (2017, methodology): “We then extracted feature vectors from the frames in the generated videos (e.g. using OpenFace).”
Sampling Matters in Deep Embedding Learning (2017, methodology): “OpenFace [2] employs offline triplet selection, so that a batch has 1/3 of images as anchors, positives, and negatives respectively.”
Adversarial Generative Nets: Neural Network Attacks on State-of-the-Art Face Recognition (2018, methodology): “ augmented images in training mini-batches. We report the performance of the networks in Table1. OF10 achieved 100% accuracy, while OF143 achieved 85.50% accuracy (comparable to Amos et al.’s finding [1]). The OpenFace DNNs were more vulnerable to naïve attacks than the VGG DNNs. For instance, OF10 ”
Motion Representations for Articulated Animation (2021, methodology): “Public reidentification networks for bodies [13] (for TaiChiHD and TED-talks) and faces [1] extract identity from reconstructed and ground truth frame pairs, then we compute the mean L2 norm of their difference across all pairs.”
DeepFakes and Beyond: A Survey of Face Manipulation and Fake Detection (2020, methodology): “Their proposed approach, named FakeSpoter, extracted as features neuron coverage behaviors of real and fake faces from deep face recognition systems (i.e., VGG-Face [37], OpenFace [38], and FaceNet [39]), and then trained a SVM for the ﬁnal classiﬁcation.”
Deep Learning for Computer Vision: A Brief Review (2018, background): “Furthermore, CNNs constitute the core of OpenFace [85], an open-source face recognition tool, which is of comparable (albeit a little lower) accuracy, is open-source, and is suitable for mobile computing, because of its smaller size and fast execution time.”
Advances in adversarial attacks and defenses in computer vision: A survey (2021, methodology): “They ascertained the susceptibility of popular model OpenFace [236] and VGG-Face [237].”
FlowSAN: Privacy-Enhancing Semi-Adversarial Networks to Confound Arbitrary Face-Based Gender Classifiers (2019, methodology): “This includes a commercial-of-the-shelf face matcher (M-COTS), which has shown state-of-the-art performance in face recognition, as well as three publicly available algorithms that provide face representation vectors: DRGAN [57], FaceNet [58], and OpenFace [59].”
Improving Smiling Detection with Race and Gender Diversity (2017, background): “Detecting and recognizing people and faces has become an increasingly feasible machine learning task due to the rise in deep learning techniques and large people-focused datasets [3, 20, 36, 27, 24, 19, 44].”
Region Attention Networks for Pose and Occlusion Robust Facial Expression Recognition (2019, methodology): “B. Implementation details In all the following experiments, we use the CNN detector and the ERT [29] based face alignment method in Dlib toolbox1 to crop and align faces, and then resize them to the 1http://dlib.net size of 224×224.”
Human Gaze Assisted Artificial Intelligence: A Review (2020, background): “Perhaps a more common but more challenging setting is that humans do not wear a camera nor an eye tracker, and the robot needs to estimate the human gaze vector by looking at their faces [Amos et al. , 2016; Saran et al. , 2018].”
Realistic Speech-Driven Facial Animation with GANs (2019, methodology): “en words. We verify the identity of the speaker using the average content distance (ACD) [36], which measures the average Euclidean distance of the still image representation, obtained using OpenFace [1], from the representation of the generated frames. The accuracy of the spoken message is measured”
Animating Arbitrary Objects via Deep Motion Transfer (2018, methodology): “Specifically, faces are aligned using the OpenFace library [1] before re-sizing each frame to 64×64 pixels.”
Facial Expression Recognition With Visual Transformers and Attentional Selective Fusion (2021, methodology): “Several face detectors like MTCNN[19] and Dlib[20] are used to detect faces in complex scenarios.”
Suppressing Uncertainties for Large-Scale Facial Expression Recognition (2020, methodology): “In face detection stage, several face detectors like MTCNN [44] and Dlib [2]) are used to locate faces in complex scenes.”
Deep Metric Learning with Hierarchical Triplet Loss (2018, background): “This is also indicated in recent literature [22,35,20,1].”
GANprintR: Improved Fakes and Evaluation of the State of the Art in Face Manipulation Detection (2019, methodology): “n detection system. Their proposed approach, named FakeSpoter, extracted as features neuron coverage behaviors of real and fake faces from deep face recognition systems (i.e., VGG-Face [21], OpenFace [22], and FaceNet [23]), and then trained a SVM for the ﬁnal classiﬁcation. The authors tested their”
Empowering Things With Intelligence: A Survey of the Progress, Challenges, and Opportunities in Artificial Intelligence of Things (2020, methodology): “For example, face recognition systems have been deployed in building entrance, railway station, and airport, enabled by cloud/fog computing [100] or edge computing [101].”